[Info-vax] History of DECSET / CMS

[Arne Vajhøj]

On 3/29/2020 11:44 PM, Andrew Shaw wrote:
> This is all good feedback guys and I can see points of relevance in
> what everyone has said to our environment. Personally I have worked
> with CMS for many many years and I have always found it appropriate
> for what we have wanted to do in our VMS environments.
> 
> The argument about moving from CMS to Git (or SVN or whatever the
> flavour du jour happens to be) is an interesting one. Where I am now,
> their source repository weapon of choice for the majority of all
> their non VMS systems happens to be git, and there are a raft of
> tools and processes in place to support that. The VMS system I help
> look after (like most I suspect) has grown steadily over a
> development period of 30+ years and still forms the core engine
> behind our central product offering. Over the more recent years a
> whole raft of supporting apps and systems have grown that do all the
> nice pretty modern stuff like web pages, containers, Kubernetes, java
> classes etc - all supporting many layered products with which we in
> the VMS world are required to interface in an increasingly complex
> application / system landscape.
> 
> The vast majority of our tech staff really don't understand this "VMS
> black box thing" and they don't really appreciate why things happen
> differently. Our dev staff are understandably mostly a generation or
> two younger than us VMS tragics and consider a lot of what we do
> "quaint".
> 
> Now I am being asked why my new code isn't in git, and I have no good
> answer. There is no reason why it can't go there, and that's where I
> will be putting it. It doesn't need to be tied to the dev processes
> of our core app. I anticipate that I will get asked soon - "Why don't
> you guys use git for your core app"? Short answer - we *could*.
> Longer answer, as a few of you have pointed out today, there would be
> non trivial work to move not only our code, but more importantly our
> build processes to a git central world. And to what end? Our build
> process works today. No, it isn't ideal, in fact its far from ideal,
> and if I was starting again I would do it completely differently,
> (but probably still within CMS, just with a differently structured
> set of libraries and cleverer use of classes etc) but I'm not. What
> are the risks with having our code in CMS today? No technical risks
> that I can see.  There is the training / staff development point
> brought up earlier which is very valid. If we want to move new devs
> into this team, that is complicated by having to get them across CMS,
> but at the end of the day our app is a VMS app - we have to get them
> across VMS too. You can't be a Dev in our world and not have an
> understanding of the VMS world. I think I'm the only guy there with a
> genuine love of, and history in, the VMS world, the others tolerate
> it and have learned how to get stuff done - and that's fine too.
> Certainly if the code side was in git or SVN or something more modern
> that would be one less technology to frighten off new hires and there
> is a plus side to that.

CMS works the same way today as it did 20 years ago.

CMS will also work the same way in 10 years as it does
today.

So there is is no hard short term risk. It will not stop
working.

But the fact that something will not stop working does not
mean that it is the right decision to continue with it.

Horses did not stop working 120 years ago, but we did switch
to cars.

If you look at the strategic perspective cost over many years,
then if your business is an evolving business then it probably
makes sense to migrate:
* less training of new people
* some time saving from more powerful tools
* less maintenance of home made extra tools
* better integration with other stuff (IDE, CI/CD etc.)

But you should plan carefully and you should probably not
see this as an isolated VCS problem.

I don't know you dev env, but let us just for fun assume that
it is:
* all work on VMS using terminal emulator
* EVE editor
* CMS with a few wrappers
* MMS for builds with a lot of wrappers

Is that what you want to be using in 5 or 10 years?

I don't think so.

But where would you go?

VSCode on PC, Git for source control and only build on VMS?

Something else? If so what??

Mentally you should start preparing for the migration by considering
current setup close to EOL.

But my gut feeling is that it may be beneficial to wait a couple
of years making the decision about future dev env.

Reason being that is not clear at this point in time where the
VMS world will be moving.

Best bet seems to be the general IT industry trend and that
means high end editor or IDE, Git and some fancy CI/CD tool.

But it is only a few months since VSI started releasing VSCode
plugins and I believe we are still missing a good CI/CD solution
(John M is running Jenkins on Linux interacting with VMS).

So there are some uncertainty.

Arne